Spiracular Control of Respiratory Water Loss in Female Alates of the Harvester Ant Pogonomyrmex Rugosus

A common feature of all ant ventilation described to date is its discontinuity (Lighton, 1988a, 1990, 1992; Lighton and Wehner, 1993; Lighton et al. 1993). Contrary to the general model of small-insect ventilation via pure diffusion, championed so convincingly by Krogh (1941) in his classic work on Cossus larvae, many small adult insects ventilate discontinuously, restricting most of their overall gas exchange (and almost all of their CO₂ release) to events spaced several minutes apart, a characteristic thought until recently to hold true only for diapausing insect pupae (see Miller, 1981; Kestler, 1985; Slama, 1988, for reviews).

Even at the risk of embracing the Panglossian paradigm (Gould and Lewontin, 1979), it is difficult to resist suggesting an obvious adaptive value for this rather extreme ventilatory behavior. The adaptive significance follows naturally from the first observed occurrence of the discontinuous ventilation cycle (DVC) in diapausing pupae. Plainly, where respiratory water loss rates must be minimized, directional selection should drive the duration of spiracular opening, during which respiratory water loss is concentrated (Kestler, 1985; Lighton, 1988b, 1992), to a minimum consistent with acceptable gas exchange rates. The widespread, if not universal, occurrence of the DVC among ants, which in many cases are exposed to extreme hygric stress, provides further supporting evidence, as does the presence of the DVC (albeit modified) in certain desert beetles (Lighton, 1991).

In fact the situation is not so simple. To explain a high intensity of selective pressure on insect ventilation and respiratory water loss rates, respiratory water loss must play a significant role in the overall water loss budget. Although numerous studies of water loss rates in insects exist in the literature, very few have directly addressed the problem of respiratory water loss rates in a rigorous manner and only three published accounts contain explicitly quantitative data in this area (Machin et al. 1991; Lighton, 1992; Hadley and Quinlan, 1993). The last two in particular clearly demonstrate that respiratory water loss rates constitute a small fraction (approximately 2–8%) of total water loss via all avenues, most notably the cuticle. But is this merely evidence for an adaptive reduction of water loss rates by means of discontinuous ventilation?

In other words, the DVC may be so efficient at reducing respiratory water loss rates in some insects that, paradoxically, those rates may be dismissed as insignificant, so that respiratory water loss and the DVC itself are considered to be unimportant. It follows that further progress in assessing the water-loss correlates of the DVC is contingent on delineating upper as well as lower limits of respiratory contributions to overall water loss rates and on doing so in a way that clearly partitions respiratory and cuticular water loss rates. If water loss rates can be followed in ‘real time’, separation of respiratory and cuticular components of water loss during discontinuous ventilation becomes practicable (Lighton, 1992) and, more importantly, peak rates of water loss during the open-spiracle phase of the DVC can be quantified, allowing an approximate estimation of water loss rates in the absence of spiracular control. It is this latter measure which, in the authors’ opinion, should be of central importance in evaluating the adaptive significance of the DVC.

We chose as our experimental animal female alates (the female reproductive caste) of the southwestern United States harvester ant Pogonomyrmex rugosus. Female alates are of particular interest because their fitness represents a ‘bottleneck’ through which all colony formation must pass (S. W. Rissing, personal communication). P. rugosus alates, like most ant alates, undergo a ‘claustral’ phase after insemination, during which they consume no food or water for several weeks until their first brood has been laid, hatched, fed, pupated and matured to the point where they can forage. At this stage the alates may be under intense selective pressure to minimize all routes of water loss. Their lowest rates of respiratory water loss therefore probably represent a genuine nadir for the species, against which their peak rates can be usefully compared.

Pogonomyrmex rugosus Emery alates were collected near Phoenix, Arizona, after monsoon rains and transported by air freight to the University of Utah, where they were kept in glass test-tubes with a moistened plug of cotton-wool at the far end. They were fed on sliced mealworms and oats, and housed at an ambient temperature of 25±2°C on a 12h:12h L:D cycle. All measurements took place in an air-conditioned laboratory at 25±1°C.

We used a modified Sable Systems TR-2 respirometry system (Sable Systems, 476 E South Temple, Salt Lake City, UT 84111) to monitor discontinuous ventilation and water loss rates, using flow-through respirometry. Dry, CO₂-free air at room O₂ concentration was pulled through the respirometer chamber (glass, 5cm³ internal volume, connected to the gas analyzer cuvette by a short length of Bev-A-Line low-permeability tubing) at a rate of 100cm³ min⁻¹ STPD, controlled by a Tylan mass flow controller. In addition to measuring CO₂ concentration in the excurrent airstream, using infrared absorbance at a wavelength of 4.26 μm, filtered by a narrow-bandpass filter and detected by a three-stage Peltier-effect-cooled detector, we employed a second alternately selected bandpass filter centered at 2.59 μm to measure infrared absorbance by H₂O (dual-wavelength infrared absorbance analysis or DWIRAA). By utilizing the same feedback-stabilized emitter and detector for both wavelengths, drift was practically eliminated and the CO₂ and H₂O signals were synchronous. After temperature correction and software linearization to account for Beer–Lambert deviations in the absorbance characteristics of CO₂ and H₂O, digitally filtered data were stored at 1s intervals. CO₂ records were then baseline-corrected and converted to rate of CO₂ production in ml h⁻¹; H₂O data were also baseline-corrected and converted to water loss rate in mgh⁻¹.

We measured mass loss directly using a Cahn C-32 ultramicrobalance, modified by suspending an extension hang-down of heat-treated Nichrome wire from its balance loop. This extension hang-down passed through a laser-aligned passage into a flow-through chamber (volume 50cm³) and was attached to a free-hanging nichrome/aluminum stirrup on which the ant was placed. The flow-through chamber was suspended by short spacers from the ceiling of a larger container (volume 1000cm³) equipped with a rubber-insulated door. Air could be pushed into the larger container through tubing connected to a 50mm diameter cylinder, through which were drilled 500 0.5mm diameter holes, and could be pulled from the small flow-through chamber through a similar number of 0.3mm diameter holes into a Lucite collar. This arrangement minimized large, irregular air currents that otherwise added significant noise to the signal from the balance. The entire apparatus was surrounded by an open-floored wooden cabinet and rested on rubber blocks directly on a concrete floor.

In use, a 250mlmin⁻¹ airstream, scrubbed of H₂O and CO₂, was pumped into the large cabinet. A sample of 10mlmin⁻¹ was pulled from the small flow-through chamber, ensuring that the ant within was exposed to dry air. The output of the ultramicrobalance was sampled by a computer at approximately 3Hz to 18-bit or 0.1 μg resolution and averages (resolution 20–50ng) were stored every 5s. Differentiating the resulting trace (in mg) against time in hours yielded mass loss rate in mg h⁻¹. Assuming a respiratory quotient (RQ) of 0.727, the molar ratio of O₂/CO₂ and a value that is appropriate for this species (Lighton and Bartholomew, 1988), we could assume that the measured mass loss was equivalent to water loss (see Lighton, 1993, for details). The mass loss measurements could only be made on decapitated ants (Lighton, 1993) because even tiny movements caused massive disruption of the mass-loss recordings. However, decapitated ants are generally very similar in terms of ventilation patterns to intact ants (Lighton, 1992; Lighton et al. 1993). The mass loss and direct water loss techniques thus served as independent checks on our respiratory water loss measurements.

Means are accompanied by standard deviations and sample sizes and are compared using Student’s t-test. Regression is by least squares, with significance testing by analysis of variance. Regressions are compared using analysis of covariance (ANCOVA).

Without exception, female Pogonomyrmex rugosus alates ventilated discontinuously. We were able to measure CO₂ emission and water loss rates simultaneously with good accuracy and resolution (Fig. 1).

To characterize ventilation and water loss characteristics, we analyzed 227 DVCs by 11 ants (mean mass 32.18±3.90mg), with approximately equal sample sizes for each ant. Summarized results are presented in Table 1.

As found in other xeric ant species (Lighton and Wehner, 1993), DVC frequency increased to accommodate elevated (Fig. 2), but burst CO₂ volume did not change with DVC frequency (r²=0.05; P>0.4). Burst CO₂ volume and burst H₂O loss were, however, tightly correlated (Fig. 3), with burst CO₂ volume explaining nearly half of burst H₂O loss variance.

Interestingly, we found little evidence of a clear fluttering-spiracle (F) phase in any of the ants we examined. This is in contrast to other xeric ants (Lighton, 1990; Lighton and Wehner, 1993). Of course, flow-through respirometry is not able to detect bulk air flow during the F phase; it is sensitive only to the diffusive component of the F phase. This component may be quite small, although generally detectable. Part of our inability to discern the F phase may have been the result of a trade-off between instrument noise and the capacity to measure both CO₂ and H₂O. Alternatively, the F phase may be highly efficient in Pogonomyrmex rugosus alates, with a negligible diffusive component. In view of the lack of a definitely measurable F phase in these ants, for analysis purposes we considered the intervals between peak CO₂ emissions (bursts) to be constricted-spiracle (C) or, more generally, interburst phases. We defined cuticular water loss rates as those measured during the interburst phase.

During the interburst phase, CO₂ was emitted at a continuous very low rate that was tightly correlated with overall (Fig. 4). From the slope of this relationship, CO₂ was lost during the interburst period at 11.6±5.6% of total ⁠, which is slightly less than the 14% of overall CO₂ lost by the arid-adapted Cataglyphis bicolor during its true C phase (Lighton and Wehner, 1993). This suggests that the F phase is indeed minimal in Pogonomyrmex rugosus.

The interburst H₂O loss rate was strongly correlated with the rate of CO₂ emission during the interburst period (Fig. 5).

Table 2 summarizes the mass loss measurements. Note that DVC frequency was significantly greater in the decapitated ants, an effect of decapitation, presumably reflecting increased (Fig. 2) not observed in decapitated Cataglyphis bicolor (Lighton et al. 1993). As with intact ants, however, the volume of water lost in the burst did not change with DVC frequency (P>0.3), and so can be directly compared to burst water loss volumes obtained by water vapor measurement in the DWIRAA system; the two values do not differ significantly (t=0.75; P>0.2). This cross-validation suggests that gravimetric and direct water loss measurements can be used interchangeably for the measurement of respiratory water loss rates in insects, subject to the stringent activity limitations of the former technique.

Comparison of interburst water loss rates is complicated by the greater DVC frequency of the decapitated ants, which, as in intact ants, significantly affected interburst mass loss rates (Fig. 6). However, the relationship between DVC frequency and interburst mass loss rates (Fig. 6 legend) can be used to predict interburst water loss rates given the mean DVC frequency of the flow-through-system ants. This value (3.02mg g⁻¹ h⁻¹ at 1.07mHz) does not differ from the mean value obtained from the flow-through-system ants (2.77±0.56mg g⁻¹ h⁻¹, Table 1; t=0.19, P>0.3). This suggests that a high cuticular water loss rate in decapitated ants is a side-effect of elevated and DVC frequency. Whether this effect is causative or correlative remains unclear.

The mean absolute cuticular water loss rate of our sample of intact Pogonomyrmex rugosus female alates (Table 1) is 0.089mg h⁻¹. Assuming a cuticular surface area of 1.04cm² (Lighton and Feener, 1989, equation 4), the cuticular permeability of female P. rugosus alates is 89 μgh⁻¹ 1.04cm⁻² 3173Pa⁻¹ of water vapor saturation deficit at 25°C=27ng H₂O cm⁻² Pa⁻¹. This is just 31% of the cuticular permeability of P. rugosus workers (87ng H₂O cm⁻² Pa⁻¹; Lighton and Feener, 1989).

The cuticular permeability of female P. rugosus alates is the lowest yet reported for any ant and, indeed, is among the lowest reported for any insect (see Edney, 1977; Machin et al. 1991). It is more than three times lower than that of worker ants of the same species, which already show the low cuticular permeabilities associated with desert arthropods (Lighton and Feener, 1989). The striking difference between alate and worker cuticular permeability in P. rugosus suggests that the phenomenon may be widespread in other ant species, especially among claustral colony founders. By extension, it is likely that such differences may be reduced or eliminated in ant species that do not undergo such rigorous isolation or do so under more favorable circumstances, for example in water-saturated soil.

Our finding that and cuticular water loss rates are correlated is somewhat surprising. This relationship is too marked to be explained by H₂O escaping in concert with CO₂ from the constricted spiracles or through the cuticle. The relationship between H₂O and CO₂ release volumes during the burst phase (Fig. 3) has a slope of 2.5±0.2 μgH₂O μl⁻¹ CO₂ while the relationship between rates of H₂O loss and CO₂ emission during the interburst phase has a slope of 29.5±3.8 μgH₂O μl⁻¹ CO₂ (Fig. 5). This strongly suggests that overall directly or indirectly modulates cuticular H₂O permeability, because changes in the interburst CO₂ emission rate (and hence overall ⁠) affect interburst H₂O loss rates tenfold more than equivalent changes while the spiracles are open, when it can be assumed that the H₂O/CO₂ loss ratio is at a maximum. Of course, the relationship between and cuticular permeability may be correlative rather than causal. Other, unmeasured factors may simultaneously affect metabolic rate (and hence ⁠) and cuticular water loss rates. Certainly, cuticular permeability has long been regarded as at least potentially labile (Treherne and Willmer, 1975; Noble-Nesbitt and Al-Shukur, 1987; but see Machin et al. 1986, 1991). Hadley and Quinlan (1993) also found evidence for alterations in cuticular permeability. Like all work in this area, our findings do not conclusively prove that cuticular permeability is under active control, but they certainly do not disprove the hypothesis. Investigating the correlation (if any) between interburst or CF water loss rates (sensuMachin et al. 1991) and metabolic rate in cockroaches may be instructive in this regard.

P. rugosus alates lose a greater percentage of total H₂O loss via respiration than either Romalea guttata grasshoppers (2–4%; Hadley and Quinlan, 1993) or other ants for which data are available (2–8%; Lighton, 1993). However, as alluded to above, the cuticular permeability of P. rugosus alates is extremely low. This greatly elevates the relative contribution of respiratory water loss to the overall water loss budget. The same may hold true of the relatively large respiratory water loss percentages reported by Machin et al. (1991; but see Hadley and Quinlan, 1993). It follows that the lower the cuticular permeability of an insect, the greater the potential selective pressure placed on its respiratory water loss characteristics. Thus, in xerophilic insects, it may be a general principle that the higher their respiratory water loss rates are in proportion to total water loss rates, the lower are their overall rates of water loss.

In certain respects, the ventilation and respiratory water loss rate characteristics of P. rugosus female alates resemble those of the xerophilic formicine ant Cataglyphis bicolor (Lighton, 1993). For example, DVC frequencies are similar (1.07±0.52 vs 1.15±0.85mHz), as are burst volumes of CO₂ (36.6±6.1 vs 35.2±4.91 μlg⁻¹). However, burst H₂O loss is lower in P. rugosus (117.6±18.5 vs 187.2±81.7 μgg⁻¹), and so is the ratio of H₂O lost to CO₂ emitted during the burst phase (3.25±0.41 vs 4.79±1.73 μgH₂Oμl⁻¹ CO₂). Furthermore, the interburst (cuticular) water loss rate is far lower (2.77±0.56 vs 6.34±5.03mg g⁻¹ h⁻¹). In general, most of the characteristics of P. rugosus appear to be better optimized to reduce both cuticular and respiratory water loss rates than are those of C. bicolor workers. This is consistent with the selective pressures that presumably operate on ant queens during claustral colony foundation.

But some data suggest a more complicated situation. Notably, the mesic ant Camponotus vicinus loses far less H₂O per burst (55.3±16.1 μgg⁻¹; Lighton, 1993) than does either Cataglyphis bicolor or P. rugosus and its CO₂ emission efficiency is greater (2.58±0.35 μgH₂O μl⁻¹ CO₂). Why, in these critical areas, is a mesic ant from montane North America paradoxically better able (in a simplistic sense at least) to reduce respiratory water loss than either of two far more xerophilic species? The requirement for more comparative data is obvious.

It has been convincingly argued (Hadley and Quinlan, 1993) that the usual reason put forward for the evolution of the DVC, namely its role in water conservation, is problematic. As they point out, their experimental organism, the mesic Eastern lubber grasshopper Romalea guttata, counterintuitively appears to ventilate discontinuously only when fully hydrated and under circumstances when the risk of desiccation is already low, such as during its inactive scotophase. Even during active ventilation, it loses only about 7% of total water loss via its respiratory system. This may, however, reflect not so much the general unimportance of the DVC or respiratory water loss, as a reduction of selective pressure on respiratory water loss caused by already high cuticular water loss rates. As Hadley and Quinlan (1993) point out, in Romalea guttata at room temperature ‘respiratory water loss is so small that it can be neglected’.

We do not consider, however, that these findings are necessarily valid for all insects. As mentioned above, respiratory water loss only becomes relevant if cuticular water loss rates are low. Our present data are unusual in that they allow quantification of both normal and maximal or near-maximal rates of respiratory water loss. This allows implicit comparison of respiratory water loss in the presence and absence of active spiracular control. (Note that non-discontinuous spiracular control regimes may entail the continuous modulation of spiracular opening near minimal levels. The water-loss correlates of such control regimes remain unclear.)

Our technique for placing an upper limit to respiratory water loss rates is very conservative. It assumes that the maximum measured H₂O emission rate during the burst phase equates to the sustained H₂O output rate that would occur if the spiracles were simply held open, i.e. if the DVC were abolished. However, the measured peak H₂O emission rate is certainly less than the sustained rate. Although H₂O is lost via diffusion at an approximately constant rate during the burst phase in ants (J. R. B. Lighton, in preparation), instrumental limitations precluded following its exact time course in either of the analytical systems employed here, in one case because of wash-out effects and in the other because of feedback time constants. Therefore, our measured rates of respiratory water loss were still pre-asymptotic when they reached their observed maximum by the end of the burst phase.

A clear picture emerges nevertheless. Cuticular H₂O loss was 2.77mg g⁻¹ h⁻¹ (Table 1) and the increase at measured peak respiratory H₂O loss was 4.91mg g⁻¹ h⁻¹, yielding a total water loss rate in the absence of the DVC of 7.68mg g⁻¹ h⁻¹, a 2.8-fold increase over non-respiratory water loss rates. Since our estimate of respiratory water loss rate is conservative, it cannot be doubted that it is under strong selective pressure in arid environments, especially in insects with low cuticular permeability.

In conclusion, our findings strongly support the water-conservation hypothesis for the role of the discontinuous ventilation cycle in insects. However, this does not mean that stringent discontinuous ventilation regimes are necessarily the norm, especially where rates of cuticular water loss are high relative to respiratory water loss rates. It is nevertheless significant that, even in such cases, the capacity for discontinuous ventilation still exists (Hadley and Quinlan, 1993), implying that it is highly conserved even when it has little demonstrable effect. The widespread occurrence of the DVC in ants strongly suggests that their cuticular water loss rates have been reduced to the point where respiratory water loss rates are under significant selective pressure.

This study was supported by the US National Science Foundation (grant BSR 9006265). J.R.B.L. thanks David Berrigan for helpful comments.